Representational Similarity Analysis (RSA) is a method used to compare patterns of responses (e.g., neural activity, computational models, or time series data) by analyzing the similarity of representational structures. It involves the following steps:

Steps in RSA:

1. Compute Representational Dissimilarity Matrices (RDMs):
   • Each row and column corresponds to a condition (e.g., stimulus, time step).
   • Each cell represents the dissimilarity (e.g., Euclidean distance, correlation distance) between two conditions.
2. Compare RDMs:
   • Compute similarity (e.g., Spearman correlation) between different RDMs.
   • Compare RDMs from different sources (e.g., brain data vs. model predictions).

Example: RSA on Synthetic Data

https://gist.github.com/viadean/772221b870788b7fdc7c9901b74bb34e

Here are the Representational Dissimilarity Matrices (RDMs) for the two synthetic sequences. The Spearman correlation between the two RDMs is displayed in the title, indicating how similar the two representations are.